docs: explain the Apple split engine

README.md

@@ -111,10 +111,14 @@ shared cache. The synthesis checkpoint will be at
 |---|---|---:|
 | PyTorch with CUDA graphs | RTX 3090 | 7.47x |
 | PyTorch with CUDA graphs | Jetson Orin Nano | 1.83x |
-| split PyTorch engine | Apple M3 Pro | 3.43x |
+| split PyTorch engine\* | Apple M3 Pro | 3.43x |
 | ONNX Runtime, CPU provider | Apple M3 Pro | 1.21x |
 | PyTorch CPU reference | Apple M3 Pro | 0.33x |
 
+\* "Split" describes device placement, not a different model or checkpoint.
+The token generator runs on the CPU while the mel and vocoder renderer runs on
+the Apple GPU through MPS. Adjacent windows can overlap across the two devices.
+
 Higher is faster, and 1.0x means real time. ONNX Runtime on the measured M3 Pro
 CPU is faster than real time. The PyTorch CPU reference path on the same machine
 is not.